Counting bicycles using computer vision

This paper describes a system for monitoring bicycle activity in sequences of gray scale images acquired by a stationary camera. The system is suitable for use in applications that aim to increase the efficiency and safety of existing traffic systems. One such application is to determine usage and congestion of bicycle path. The output of the system is a count of the number of bicycles detected in the image sequence. The system is model-based in the sense that it uses a simple model of two circular objects separated by a relatively known distance. Our system uses four levels of abstraction: raw images, blobs, edge images, and the bicycle model. The system was implemented on a dual Pentium computer equipped with a Matrox imaging board and achieved a peak performance of 8 frames per second. Experimental results based on outdoor scenes have shown promising results for a variety of weather conditions.